Limited Effects of a Keystone Species: Trends of Sea Otters and Kelp Forests at the Semichi Islands, Alaska

MARINE ECOLOGY PROGRESS SERIES Vol. 199: 271-280,2000 Published June 26 Mar Ecol Prog Ser I

Limited effects of a keystone species: trends of sea otters and kelp forests at the Semichi Islands, Alaska

Brenda Konar*

U.S. Geological Survey, University of California, Santa Cruz, California, USA

ABSTRACT: Sea otters are well known as a keystone species because of their ability to transform sea urchin-dominated communities into kelp-dominated communities by preying on sea urchins and thus reducing the intensity of herbivory. After being locally extinct for more than a century, sea otters re-colonized the Sernichi Islands in the Aleutian Archipelago, Alaska in the early 1990s. Here, otter populations increased to about 400 individuals by 1994, but rapidly declined to about 100 by 1997. Roughly 7 yr after initial otter re-colonization, there were only marginal changes in sea urchm biomass, mean maximum test size, and kelp density. These small changes may be the first steps in the cascahng effects on community structure typically found with the invasion of a keystone species. However, no wholesale change in community structure occurred following re-colonization and growth of the sea otter population. Instead, this study describes a transition state and identifies factors such as keystone species density and residence time that can be important in dictating the degree to which otter effects are manifested.

KEY WORDS: Community structure . Trophic interactions . Urchin barrens . Enhydra lutris . Strongy- locentrotuspolyacanthus . Alaria fistulosa

INTRODUCTION proportionately large relative to its abundance, and thus keystones are usually rare (Power et al. 1996). The various species that belong to an ecological Dominant species are cominonly assumed to be of community can assume diverse roles within the food primary importance according to the bottom-up view web and conlpetitive hierarchy, and will always rank of trophic interactions, which considers lower-level very differently in their importance in structuring the producers to have a primary role in structuring com- community (Paine 1992, Mills et al. 1993, McCann et munities (see Hunter & Price 1992, Power 1992 for al. 1998).While there is clearly a gradient in the impor- reviews). This, however, is not always the case. A top- tance of species, a small number of species is often down view of trophic interactions may also charac- easily identified as occupying positions of particularly terize dominants as playing a controlling role in com- strong influence on community structure. These spe- munity structure. This is illustrated in systems where cies often fill 1 of 2 mutually exclusive roles: dominant herbivores or omnivores are not limited by predation. or keystone. Dominant species are ones in which den- In this case, animal abundances can increase, such that sities or total biomass are very high, and for which these organisms become important dominants, sup- community influence is a direct result of abundance. A planting the controlling role of plants. Classic exam- keystone species is defined as one whose effect is dis- ples of such dominant species are sea urchins (Andrew & Underwood 1993) and mussels (Paine 1974). In some 'Present address: School of Fisheries and Ocean Sciences, systems, sea urchins can become so abundant that they University of Alaska Fairbanks, PO Box 757220, Fairbanks, overgraze attached plants in kelp forests, leaving what Alaska 99775-7220. USA. E-md: [email protected] are commonly termed 'barren grounds' (areas devoid

O Inter-Research 2000 Resale of full article not permitted 232 Mar Ecol Prog Ser

of macroalgae; Arnold 1976, Chapman 1981, Schiel fects exerted by a keystone species on community 1982, Dean et al. 1984, Harrold & Pearse 1987, Watan- structure do not depend on high densities, but the ex- abe & Harrold 1991), while mussels can form mono- tent to which community-wide effects of keystone spe- cultures by outcompeting other sessile organisms for cies depend on the attainment of some critical density space (Paine 1974, Robles et al. 1995). has not been carefully evaluated. Here these issues are Members of a second group of important species addressed by examining the keystone role of sea otters have been labeled as keystones (sensu Paine 1969).By Enhydra lutris in kelp forest communities in the Semi- current definition, keystone species are comparatively chi Islands of the Aleutian Archipelago, Alaska. rare in the communities in whch they reside (Power et The Semichi lsland group consists of 3 islands, al. 1996). Keystone species have been identified in Shemya, Nizki and Alaid (Fig. 1). Shemya Island numerous ecosystems and at all trophic levels (see (52"43'N, 174"07'E) is the furthest east and largest of Mills et al. 1993, Power et al. 1996 for reviews). In these 3 islands. The island has a shoreline of 22 km. communities controlled by top-down forces, the key- Nizki and Alaid are slightly smaller than Shemya and stones are often apex predators. These predators con- are attached to each other by a sand bar at low tide. trol their prey (often herbivores), which otherwise act There is approximately a 1.5 km wide channel as dominants, thereby exerting strong effects on com- between the islands of Shemya and Nizki. All 3 of munity structure. The influence of keystone species is these islands are exposed directly to swell and wave best seen by observing systems in which the key- action from the Bering Sea to the north and the Pacific stone's population status is in flux, either through pur- Ocean to the south. poseful or fortuitous experiments. In this situation, a rise or fall in the keystone population abundance can cause cascading effects on its prey, their food species, THE ALEUTIAN NEARSHORE ECOSYSTEM and ultimately on the entire food chain. (Fretwell 1987). There has been a tendency in the ecological litera- Sea otter population status ture to apply the labels 'dominant' or 'keystone' to species wherever they occur. In fact, the evidence from Sea otters were once abundantly distributed in nature indicates that these roles are highly context coastal waters across the Pacific rim from Japan to cen- dependent (Foster 1990, Menge et al. 1994, Dean et al. tral Baja, California (Kenyon 1969).By the early 1900s, 2000). The ability of a given species to exert a strong the Pacific maritime fur trade had caused the local influence on a community structure (as either a domi- extinction of sea otters in most of the Aleutian Islands. nant or a keystone) is dependent upon community Following their protection by the International Fur composition and can be particularly sensitive to condi- Seal Treaty in 1911, sea otters remnant at Rat Island tions. For example, sea urchins may be dominant and recolonized the Rat Island group, and apparently capable of overgrazing attached macroalgae in one area while in another area or at a different time they have no dominant effect because of variation in factors such as food availability (Harrold & Reed 1985), water movement (Ebeling et al. 1985), or population density (Andrew & Underwood 1993). Similar context- 60" 'lands dependent variation in the keystone effects of seastar (Menge et al. 1994) and lobster (Robles 1997) preda- 'S* ,....a tion on intertidal mussels has been demonstrated for 160" 140" changes in wave exposure and sedimentation. With changes in these conditions, a potentially important ---$,.'g species may quickly change from a minor player to one = E actively controlling community structure, and vice . versa. While such context dependence is recognized for Pacific Ocean Nizki 0. C some species, few studies have quantitatively ad- Shemya dressed the factors that may limit the effect of a keystone species on community structure (except see Dean et al. 2000). In particular, it is not clear how sensitive the Fig. 1. Map of the Semichi Islands showing the location of influences of keystone species are to internal factors Alaid, Nizlu, and Shemya with dive iocations (0). inset: map such as the amount of time a species has been present of Alaska showing the location of the Semichi Islands and in a community or to its density. By definition, the ef- neighboring island groups Konar: Effects of sea 01 tters as a keystone species 273

reached carrying capacity throughout this area by the cluding the rock jingle Pododesmus macroschisma; De- 1950s. The Near Island group, about 225 km further shayes), chitons (including Katharina tunicata; Wood, west, did not become re-occupied until the mid-1960s, and Cryptochiton stelleri; Middendorf), limpets, crabs when a small number of sea otters re-colonized Attu and seastars are all common inhabitants in this subtidal Island. The sea otter population at Attu grew rapidly community. All of the above are known prey items of the through the 1970s and 1980s (Estes 1990). Another sea otter in the Aleutians particularly when sea urchins Near Island, Agattu, was recolonized in the early are small or scarce (Estes et al. 1981, VanBlaricom & 1980s. The Semichi Islands, approximately 30 km to Estes 1988). the northeast, were recolonized in the early 1990s. Rocky subtidal habitats are patchy around the Semi- In the Aleutian Islands, sea otters are considered chi Islands. This leads to a patch mosaic of kelp forests keystone species because of their influence as preda- and barren grounds. Course-grained sand surrounds tors in kelp forest ecosystems (Estes & Palmisano 1974, most of these rocky outcroppings. No macroinverte- Estes et al. 1978, Estes & Duggins 1995), where they brates were found in these sandy areas (author's pers. can eat approximately 15 to 20% of their body mass obs.) because of the overall exposure to swell and cur- daily (Kenyon 1969). Sea urchins Strongylocentrotus rents. Also, no sea otters were ever observed foraging polyacanthus are a preferred prey of sea otters (Estes in these areas, probably due to the lack of inverte- et al. 1981), and at high densities are capable of com- brates (author's pers. obs.). pletely deforesting established kelp beds (Harrold & Pearse 1987).While several other species consume sea urchins in the Aleutian Islands (e.g., sea stars and sea MATERIAL AND METHODS ducks), none of these appear to be capable of limiting urchin populations. Estes & Duggins (1995) showed Sea otter population changes. Population counts of that on islands where sea otter density was high, sea sea otters at the Semichi Islands were done during urchin test diameters were relatively small and bio- 4 years in a 10 yr period (1987-1997) to determine mass low while algal cover was high. Conversely, on their status and trends. Surveys were done in a single islands where sea otters were rare or absent, sea day by systematically searching the entire coastline of urchin test diameters were significantly larger and bio- Alaid, Nizki, and Shemya Islands from small boats by mass higher while algal cover was low. at least 2 people using binoculars. This survey method was chosen because ground-truthing at other Aleutian Islands showed boat surveys to be efficient and accu- Subtidal community structure rate (Estes 1977, 1990; except see Udevitz et al. 1995 for Prince William Sound). These searches were usu- Kelp forest communities in the western Aleutian ally done when sea states were very calm (low to mod- archipelago contain a rich diversity of algal species erate swell and no wind chop). One survey was done in (Lebednick & Palmisano 1977). Surface canopies are both 1987 and 1994 and 2 surveys were done in each of formed by the large annual kelp Alaria fistulosa (Pos- 1995 and 1997. Within-year means were calculated for tels et Ruprecht). The most common understory kelps years in which multiple surveys were done. are the perennial brown algae Laminaria dentigera Population counts, such as these, are naturally going (Kjellman), L. yezoensis (Miyabe), Agarum cribrosum to be biased low as some sea otters will always be (Bory) and Thalassiophyllum clathrus (Gmelin; Postels missed by any method employed. However, the overall et Ruprecht). Numerous species of foliose red algae trends in population status will hold. Comparison of occur beneath the understory kelp canopy. In areas these population estimates to other areas should be where the perennial brown kelps dominate, sea urch- done with caution, as different survey methods and ins Strongylocentrotuspolyacanthus are generally rare locations will give varying degrees of bias. and small in size. Alternatively, in urchin-dominated Benthic community structure. To measure effects of zones, grazing has completely eliminated foliose algal changing sea otter densities at the Semichi Islands, cover, forming barren grounds. In these areas, encrust- benthic community structure was quantified both being coralline algae (mostly Clathromoxphum spp. and fore sea otters became re-established (sometime in the Lithothamnium spp.) are the primary cover except for late 1980s), several years later (1994) after otter num- patches of the prostrate green alga Codium setchellii bers had increased, and then in 1997 as otter densities (Setchell et Gardner). were declining. To quantify community structure, algal Kelp forests provide habitat for a diverse array of fish density and urchin biomass were examined at the same and invertebrates. Fish (including the Irish lords Hemi- randomly chosen sites around eachisland in 1987,1994 lepidotus spp., greenlings Hexagrammos spp., and the and 1997 (for methods of site determination and sam- rock sole Lepidopsetta bilineata; Ayres), bivalves (in- pling details, see Estes & Duggins 1995).Depending on 274 Mar Ecol Prog Ser 199: 271-280, 2000

the year, between 28 and 34 sites were surveyed at 7 m Alaid water depth and either 8 or 9 sites at 13 m water depth. 1 Nizki Kelp densities were determined by counting the stipes Shemya in twenty 0.25 m2quadrats at each site/depth combina- \ -0- total otters tion. Placement of each individual quadrat was determined by a random number of kicks along a transect line. To determine sea urchin biomass, individuals were counted and collected from another set of randomly placed 0.25 m2 quadrats at each site. Urchins from any given site were sampled until at least 200 individuals were collected or 20 quadrats were surveyed. These urchins were taken to the surface, where their test diameters were measured to the nearest millirneter. Bio- mass was determined using these measurements and a linear model as in Estes et al. (1978).An overall mean 1987 1994 1995 1997 maximum test diameter for each depth and year was n=l n=l n=2 n=2 determined by averaging'the diameters of the largest year urchin that was collected and 'measured in'the previ- ously described quadrats at each site. Fig. 2. Mean number of sea otters (21 SE for 1995 and 1997) at Sea otter foraging behavior. To determine the Alaid, Nizki, Shemya, and all islands combined for 1987, 1994, 1995, and 1997. Number of surveys conducted at each primary food of sea otters, shore-based observations island is shown below the year of feeding animals were undertaken periodically between June 1995 and June 1997 using a high powered Questar spotting scope. Following each dive, the type of 2.3 otters per km of shoreline. In 1997, 118 + 10.0 and number of the prey daptured were recorded. A otters were counted with a density of 2.0 otters per km single foraging bout was defined as an unbroken se- of shoreline. Based on these numbers, the total Semichi quence of dives made by 1 otter. The otters were otter population declined by 65% between 1994 and watched until they stopped feeding, moved too far 1995 and by an additional 16 % between 1995 and away to be observed or were joined by other otters 1997. In 1994 and 1995, otters were most abundant at making it difficult to distinguish them. To supplement Alaid, whereas in 1997, they were found to be equally these data, sea otter scats were examined on Shemya abundant across all 3 islands. Island in the rocky intertidal from November 1995 to May 1997. On this island, otters haul out as a group during low tide at very specific sites (Estes et al. 1999). Benthic community structure A total of 704 scats were examined during low tides after the otters had been hauled out of the water for at During all years, sea urchin biomass appeared some- least approximately 2 h. To assure this minimum sur- what higher at 7 m water depth than at 13 m, although face time, once hauled. out otters were located, the haul a 2-way ANOVA showed that this was not statistically out sites were revisited a minimum of 2 h after the ini- significant and there was no interaction (Fig. 3, tial sighting. The scats were dissected in the field and Table 1). Also, there was no significant difference in the percent of each food item in the scat (by volume) sea urchin biomass between years (Fig. 3, Table 1).I was visually estimated. am confident that the overall lack of differences found both between years and depths is real (power = 0.57 and 0.99 for year and depth respectively). The differ- RESULTS Table 1. Two-way ANOVA of the effect of year (1987, 1994, Sea otter population changes and 1997) and water depth (7 and'l3 m) on sea urchin biomass at randomly selected sites at the Semichi Islands The sea otter population in the Semichi Islands var- ied greatly between 1987 and 1997 (Fig. 2). From sin- Source df MS F-ratio P gle surveys of all 3 islands, 1 otter was counted in 1987 and 390 (approximately 6.5 otters per km of shoreline) Year 2 105798 1.46 0.25 Depth 1 157346 2.11 0.15 in 1994. Two surveys in both 1995 and 1997 docu- - - rear X Depth 2 6535 0.09 3.92 mented a decline in the otter population. In 1995, 134.5 Error 115 74544 + 28.2 (mean i 1 SE) otters were counted with a density Konar: Effects of sea otters as a keystone species 27 5

50 V) 1994: n=30 sites a, 96 sites 94-97 1997: n=33 sites .% 40 C 0 a, 30 m0 20 a, l0 a 0......

n=8 sites 1987 1994 1997 n=8 sites n=9 sites year

Fig. 3. Mean sea urchin biomass per 0.25 m2 (+l SE) at the Semichi Islands for 1987, 1994, and 1997. Number of sites sampled is shown above standard error bars. No significant differences were found among years or water depths using a 2-way ANOVA ence in biomass for each pair of sampling dates at each site was also calculated to determine the percentage of sites that had either an increase or decrease in biomass over time. While biomass at some sites increased and others decreased over each time interval (Fig. 4), the overall trend was a decline in biomass between 1987 difference in sea urchin mean biomass (g) and 1994. In contrast, the trend for urchin biomass Fig. 4. Change in sea urchin biomass for each of the sites over between 1994 and 1997 was a slight but non-signifi- time. Net losses and gains were calculated from 1987 to 1994 cant increase at 7 m and no discernible trend at 13 m. and from 1994 to 1997 for all the Semichi Island sites com- The range in differences in the change of urchin bio- bined at 7 and 13 m mass at 13 m was less than at 7 m (-900 to +g00 compared to -1500 to + 1200); however, there were no significant differences between time interval distributions in sites that had varying changes in urchin biomass (1987 to 1994 and 1997 to 1997) within either depth (2-sample Kolmogorov-Srnirnov test: for 7 m, p = 0.163 and for 13 m, p = 0.801). The mean maximum test diameter of sea urchins changed significantly with year but not depth and there was a non-significant interaction (Fig. 5, Table 2). Iam confident that the lack of significant dif-

Table 2. Two-way ANOVA of the effect of year (1987, 1994, and 1997) and water depth (7 and 13 m) on maximum mean sea urchin test diameters at randomly selected sites at the Sernichi Islands year Source df MS F-ratio P Fig. 5. Mean maximum sea urchin test diameter (+lSE) at the Semichi Islands for 1987, 1994, and 1997. The number of sites Year 2 847 8.74 0.00 sampled is shown above the standard error bars. There were Depth 1 175 1.81 0.18 no significant differences between depths. Similar letters Year X Depth 2 24 0.25 0.78 above the standard error bars denote non-significant differ- Error 111 97 ences among years (A-B) using a post-hoc Scheffe F-test, p < 0.05 after a 2-way ANOVA percent individuals in size group

E 2 a5 %pr m

;GP $5raZ Q.TP %:.a ,: mW m Li 5 %am pwsr ;2 m0 10,

L",%-. E a. ;; Et. ",S g Gg W - Konar: Effects of sea otters as a keystone species 277

were found in the scats that also were not seen being consumed by otters during the foraging observations.

DISCUSSION

Some species can play an important role in structuring their communities. Their influences can vary among systems (Mills et al. 1993) and may be context- dependent within a system (Foster 1990, Menge et al. 1994, Robles 1997). This study of the effects of sea otters on subtidal communities at the Aleutian Islands is one of the first to present quantitative data suggest-- - 1987 1994 1997 ing that internal variables (density and residence time) year may be influencing the effects or mediating the time required to see effects of keystone species (see also Fig. 7. Mean total number of kelp stipes per 0.25 m2 (+lSE) at Dean et al. 2000). the Semichi Islands for 1987, 1994, and 1997. Number of sites The pattern involving keystone species in the near- sampled is shown above standard error bars. There were no significant differences between depths. Similar letters above shore subtidal community at the Semichi lslands dif- the standard error bars denote non-significant differences fered from those of most other well-studied areas in the among years (A-B) using a post-hoc Scheffe F-test, p < 0.05 northern Pacific. In particular, a rapid, wholesale after a 2-way ANOVA change in community structure (from urchin-dominated to kelp-dominated) was not observed with the re-establishment of the potential keystone species, the ance in kelp density among plots. No significant differ- sea otter. This rapid rate of change is exemplified in ences were found between depths (power = 0.54) and some areas around Vancouver Island, British Colurn- there was no interaction between year and depth. In bia, where a change from an urchin-dominated to general, overall kelp densities were lowest in 1987 algal-dominated community occurred in under 6 mo (when otters were absent),highest in 1994 (when otter following the re-establishment of sea otters (Watson numbers were highest), and somewhat lower in 1997, 1993). When comparing the Semichi Islands to other when otter numbers were in decline (Fig. 7). Scheffe's study sites in the Aleutian Archipelago (Estes & post-hoc contrasts revealed that kelp density was lower Palmisano 1974, Estes et al. 1978, Estes & Duggins in 1987 than in either 1994 or 1997 and that the 2 later 1995), southeastern Alaska (Estes & Duggins 1995), years did not differ significantly (Fig. 7). and British Columbia (Watson 1993), changes in the nearshore community structure at the Semichi Islands were limited. At the Semichi Islands, fluctuations in Sea otter foraging behavior sea otter numbers were observed and were found to correspond with small changes in sea urchin biomass, In total, 1518 sea otter foraging dives were observed, mean maximum test diameters, and kelp stipe densi- spread across 112 foraging bouts. Most of these dives ties. However, these changes were limited when com- were observed on the north side of Shemya Island pared to other areas. An example of this is illustrated because this was where most otters were found. Both with kelp stipes. Although kelp stipe densities in- direct foraging observations and scat data on Shemya creased between 1987 and 1994 (Fig. ?), peak stipe Island showed that sea urchins were the otters' pri- densities achieved in the presence of otters were low mary food item (83.8% of foraging items and 94.7 % of when compared to other islands with abundant sea scat items).Other forage items included fish (Irish lords otters (i.e.where kelp stipe densities can reach over 15 Hemilepidotus spp., the rock sole Lepidopsetta bilin- per 0.25 m2 compared to a high of 3.5 per 0.25 m2 in eata and one smooth lumpsucker Aptocyclus ventrico- this study; Estes & Duggins 1995). sus), bivalves (including the rock jingle Pododesmus The data from the Semichi Islands suggest that the macroschisma), chitons (including Katharina tunicata mere presence of a potentially keystone species is not and Cryptochiton stelleri),limpets, crabs and seastars. enough to allow that species to fulfill the role of a classic I observed otters eating seastars, which lack hard keystone. This is supported by another study in Prince parts, but did not find seastar remnants in the scats. I William Sound (Dean et al. 2000) that showed that the ef- also did not find chitons and limpets, which do have fects of reduced sea otter density on size distribution is calcareous exoskeletons, in the scats. No prey items predictable but that sea urchin biomass and the cascad- 278 Mar Ecol Prog Ser

ing effects on the kelp community are less predictable. In food-limited. Second, sea urchin biomass remained other studies, a lack of a complete community modifica- greatest in shallower water (Fig. 3). In other locations tion by a keystone species has been attributed to various where sea otters have reached carrying capacity, sea environmental factors, such as sedimentation and wave urchin densities are greatly reduced in shallower exposure (Menge et al. 1994, Robles 1997).In this study, depths (7 vs 13 m; Estes et al. 1981) because foraging the lack of community transformation was apparently costs are less in shallower water and otters preferen- due to non-environmental factors. One such factor that tially choose those depths. Based on the observation appears to be influencing this effect is keystone density. that the primary food item of sea otters was sea urchins Although the definition of a keystone states that their in- during this period (direct observation of feeding otters fluences are disproportionately large relative to their and examination of scats),it appears that the decline in abundance, the data from this study suggest that the keystone density was not related to food limitation, keystone phenomenon may require a threshold density. implying that carrying capacity was never reached. Although clear density dependence in the functional role Similar to the Semichi Islands, there was another of keystone species has not before been documented, it island in the western Aleutians (Attu) that did not has been shown that some dominant species must reach exhibit a complete change from urchin-dominated to a threshold density in order to cause a dramatic change kelp-dominated after sea otters became re-established in community structure (Andrew & Underwood 1993).In but before they reached their carrying capacity (Estes Australia, when various densities of sea urchins were & Duggins 1995). There were 2 notable differences added to kelp stands, only the highest densities were between Attu and the Semichis. At Attu, although sea capable of removing all the macroalgal cover. At the urchin densities were high, mean maximum test diam- Semichi Islands, peak sea otter densities were approx- eter was small. On this island, the sea otter population imately 6.5 km-' in comparison to numbers for a neigh- was increasing between the 1970s and early 1990, and boring island, Amchitka, where densities reached had not yet reached its carrying capacity; however, sea 32 km-' in 1992 (J.A. Estes unpubl. data, using similar otters were able to reduce the mean maximum size of techniques to reduce sampling bias). Thus, it may be that the urchins (Estes & Duggins 1995). This differs from otter density at the Semichis was not high enough to the Semichis in that mean maximum test diameter at cause a dramatic change in the biomass of thedominant the Semichis was still large after the incomplete recov- sea urchins. ery of the otters and despite up to 5 yr of otter resi- At other islands in the Aleutians, sea otters have dence. This implies that time of occupation (20 yr on been shown to increase until food became limiting Attu vs 5 yr at the Semichis) might be influencing sea (Estes & Duggins 1995).At the Semichi Islands, follow- urchin populations. In another study in Prince William ing an initial increase in sea otter population size, a Sound Alaska, sea otters were absent for 9 yr but decline was observed between 1994 and 1997, when sea urchin biomass and kelp densities remained my study was concluded. These relatively low num- largely unchanged (Dean et al. 2000). The other bers of sea otters never reached their theoretical carry- noticeable difference between Attu and the Semichi ing capacity (based on remaining food availability) Islands was that Attu had a fairly consistent recruit- before their abundances began to decline. When sea ment of sea urchins (Estes & Duggins 1995). These otters are food-limited they typically maintain a large small urchins are capable of grazing much of the algal prey diversity in their diet due to the scarcity of their cover and probably asslsted in maintaining the barren preferred food sources (Estes et al. 1981). On Shemya state of the community. This did not occur during my Island, otters showed low diversity in their diet, feed- study at the Semichi Islands as small urchins were rare ing primanly on sea urchins (83.8% of foraging obser- between 1995 and 1997 (author's pers. obs.). Many vations and 95 % of scat data). At Adak Island, where small urchins were found at these islands in 1987 sea otters had reached carrying capacity, foraging data (Estes & Duggins 1995), so it seems that urchln recruit- showed greater diversity in sea otter diet, including ment at the Semichi Islands may be episodic. bivalves, fish, crabs and worms, although sea urchins It appears that the abundance of sea otters in the remained important prey (33%; Tinker & Estes 1996). Semichi Islands began to decline before they reached Other data also indicate that sea otters were not food- numbers sufficient to reduce sea urchin densities and limited at Shemya Island. First, sea urchin mean maxi- relax grazing pressures on the algal community. This . mum test diameter and biomass remained relatively situation is mirrored at other islands in the Aleutian high even in 1994, at peak otter abundances. Previous Archipelago where sea otter abundances have also work, which shows that otters preferentially feed on declined through the 1990s (Estes et al. 1998). One larger sea urchins (VanBlaricom & Estes 1988), indi- hypothesis for this reduction in the sea otter population cates that mean maximum sea urchin test diameters is the possible addition of a higher trophic level key- should have greatly decreased had otters become stone predator, the killer whale. Estes et al. (1998) Konar: Effects of sea otters as a keystone species 279

found that the decline in the sea otter population on Chapman ARO (1981) Stability of sea urchin dominated bar- Adak Island was coincident with the decline of kelp ren grounds following destructive grazing of kelp in St. beds. Areas that contained dense kelp beds are now Margaret's Bay, Eastern Canada. Mar Biol 62:307-311 Dean TA, Schroeter SC, Dixon JD (1984) Effects of grazing by devoid of most foliose macroalgae. At the Semichi two species of sea urchins (Strongylocentrotus francis- Islands, there was a 70 % decline in the sea otter popu- canus and Lytechnus anamesus) on recruitment and sur- lation between 1994 and 1997. Surprising, only 1 otter vival of two species of kelp (Macrocystis pyrifera and carcass was found on the beach during the time of the Pterygophora californica). Mar Biol78:301-313 Dean TA, Bodkin JL, Jewett SC, Monson DH, Jung D (2000) decline (Estes et al. 1999). The cause of reduction in Changes in sea urchins and kelp following a reduction in sea otters at the Semichi Islands is unknown but may sea otter density as a result of the Exxon Valdez oil spill. also be a consequence of enhanced killer whale pre- Mar Ecol Prog Ser 199:281-291 dation. Although killer whales were never observed Ebeling AW, Laur DR, Rowley RJ (1985) Severe storm distur- feeding on sea otters at Shemya Island, they were bances and reversal of community structure in a southern California kelp forest. Mar Biol84:287-294 occasionally sighted in the nearshore environment Estes JA (1977) Population estimates and feeding behavior of (author's pers. obs). This reduction in sea otters is prob- sea otters. In: Merritt ML, Fuller RJ (eds) The environment ably not due to otter emigration because females with of Amchitka Island. Alaska. US Energy Research and pups were observed every year, suggesting year round Development Administration, Springfield, VA, p 526 Estes JA (1990) Growth and equilibrium in sea otter popula- occupancy as opposed to overall population move- tion~.J Anim Ecol59:385-401 ments (Estes et al. 1999). Estes JA, Duggins DO (1995) Sea otters and kelp forests in In this study, the density and occupation time of a Alaska: generality and variation in a community ecologi- keystone species were shown to be possible explana- cal paradigm. Ecol Monogr 65:75-100 tions for the lack of an overall community change. At Estes JA, Palmisano JF (1974) Sea otters: their role in structuring nearshore communities. Science 185:1058-1060 the Semichi Islands, sea otters were unable to generate Estes JA, Smith NS, Palmisano JF (1978) Sea otter predation a substantial cumulative effect on the nearshore ben- and community organization in the western Aleutian thic community after 5 yr of residence. In particular, Islands, Alaska. Ecology 59:822-833 they were unable to achieve large reductions in the Estes JA, Jameson RJ, Johnson AM (1981) Food selection and some foraging tactics of sea otters. In: Chapman JA, overall mean maximum test size and shallow-water Pursley D (eds)Worldwide Fubearer Conference Proceed- densities of the sea urchins. Although keystone species ings, Vol 1. University of Maryland Press, Baltimore, MD, have been defined as those whose effect is dispropor- p 606-64 1 tionately large relative to their abundance (Power et al. Estes JA, Tinker MT, Williams TM, Doak DF (1998) Wer 1996), it appears that a threshold of abundance and whale predation on sea otters linking oceanic and nearshore ecosystems. Science 282:473-476 time must be reached to result in strong cascading Estes JA, Konar BK, Tinker MT (1999) Sea otter population community effects. biology and subtidal community ecology at Shemya Island, Alaska. Final report for Department of Defense Legacy Project Nos. 9401280 and 9510014 Acknowledgements. This project was funded by the Depart- Foster MS (1990) Organization of macroalgal assemblages in ment of Defense, Legacy Resource Management Program, the Northeast Pacific: the assumption of homogeneity and the USGS-Biological Resources Division, and the National the illusion of generality. Hydrobiologia 192:21-33 Science Foundation. Much of the data is from on-going stud- Fretwell SD (1987) Food chain dynamics: a central theory in ies by J. Estes. I would like to thank J. Estes, D. Doak, P. Rai- ecology? Oikos 50:291-301 mondi, E. Joules, S. Kohin, and D. Monson for offering com- Harrold C, Pearse JS (1987) The ecological role of echino- ments that greatly improved a previous version of this derms in kelp forests. In: Jangoux M, Lawrence JM (eds) manuscript. I would also like to thank my primary field assis- Echinoderm studies. AA Balkema, Rotterdam, p 137-233 tants: C. Roberts and C. Dominic. Many thanks also go to M. Harrold C, Reed DC (1985) Food availability, sea urchin Kenner (U. C. Santa Cruz), G. Augustine and J. Copeland grazing, and kelp forest community structure. Ecology 66: (DoD), and J. Bodkin, D. Monson, and G. Esslinger (USGS- 1160-1 169 Biological Resources Division, Anchorage) for informal sup- Hunter MD, Price PW (1992) Playing chutes and ladders: het- port. Thanks also need to go to the US Fish and Wildlife Ser- erogeneity and the relative roles of bottom-up and top- vice-Alaska Maritime Refuge and the US Coast Guard for down forces in natural communities. Ecology 73:724-732 logistical support. Kenyon KW (1969) The sea otter in the eastern Pacific Ocean. N Am Fauna 68:l-352 Lebednick PA, Palmisano JF (1977) Ecology of marine algae. LITERATURE CITED In: Merritt ML, Fuller RJ (eds) The environment of Am- chitka Island, Alaska, US Energy Research and Develop- Andrew NL, Underwood AJ (1993) Density-dependent forag- ment Administration, Springfield, VA, p 353-393 ing in the sea urchin Centrostephanus rodgersii on shal- McCann K, Hastings A, Huxel GR (1998) Weak trophic inter- low subtidal reefs in New South Wales, Australia. Mar actions and the balance of nature. Nature 395:794-798 Ecol Prog Ser 99:89-98 Menge BA, Berlow EL, Blanchette CA, Navarrette SA, Yam- Arnold DC (1976) Local denudation of the sublittoral fringe by ada SB (1994) The keystone species concept: variation in the green sea urchin, Strongylocentrotus droebachiensis interaction strength in a rocky intertidal habitat. Ecol (0.F. Muller). Can Field-Nat 90: 186-187 Monogr 64:249-286 280 Mar Ecol Prog Ser 199: 271-280, 2000

Mills SL, Soule ME, Doak DF (1993) The keystone-species Schiel DR (1982) Selective feeding by the echinoid, Evichinus concept in ecology and conservation. Bioscience 43: chloroticus and the removal of plants from subtidal algal 219-224 stands in northern New Zealand. Oecologia 54:379-388 Paine RT (1969) A note on trophic complexity and community Tinker T, Estes JA (1996) The population ecology of sea otters stability. Am Nat 103:91-93 at Adak Island, Alaska. Final Report (12/96) Natural Paine RT (1974) Intertidal community structure: experimental Resources Section/Code 23/KL Engineering Field Activ- studies on the relationship between a dominant competi- ity, NW Naval Facilities Engineering Command, Poulsbo, tor and its principal predator. Oecologia 15:93-120 Washington Paine RT (1992) Food-web analysis through field rneasure- Udevitz MS, Bodkin JL, Costa DP (1995) Detection of sea ment of per capita interaction strength. Nature 355:73-75 otters in boat-based surveys of Prince William Sound, Power ME (1992) Top-down and bottom-up forces in food Alaska. Mar Marnm Sci 11:59-71 webs: do plants have primacy? Ecology 73:733-746 VanBlaricom GR, Estes JA (1988) The community ecology of Power ME, Tilman D, Estes JA, Menge BA, Bond WJ, Mills LS, Sea Otters. Springer-Verlag, Berlin Daily G, Castilla JC, Lubchenco J, Paine RT (1996) Chal- Watanabe JM, Harrold C (1991) Destructive grazing by sea lenges in the quest for keystones. Bioscience 46: 609-620 urchins Strongylocentrotus spp. in a central California Robles CD (1997) Changing recruitment in constant species kelp forest: potential roles for recruitment, depth, and pre- assemblages: implications for predation theory in inter- dation. Mar Ecol Prog Ser 71:125-141 tidal communities. Ecology 78: 1400-1414 Watson J (1993) The effects of sea otter (Enhydra lutris) forag- Robles CD, Sherwood-Stephens R, Alvarado M (1995) Re- ing on shallow rocky communities off northwestern Van- sponses of a key intertidal predator to varying recruitment couver Island, British Columbia. PhD thesis, University of of its prey. Ecology 76:565-579 California, Santa Cruz, CA

Editorial responsibdity: Charles Peterson (Contributing Submitted: June 28, 1999; Accepted: January 11, 2000 Editor), Morehead City, North Carolina, USA Proofs received from author(s): June 5, 2000